micro-RNA Vesicles

Устанавливаем необходимые библиотки

library(tidyverse)
library(DESeq2)
library(pheatmap)
library(RColorBrewer)
library(clusterProfiler)
library(biomaRt)
library(org.Hs.eg.db)
library(EnhancedVolcano)
library(GenomicRanges)
library(msigdbr)
library(multiMiR)
library(miRBaseConverter)
library(enrichplot)
library(vsn)
library(rvest)
library(patchwork)
library(dbplyr)

Импортируем данные

coldata <- read_tsv("data/phenotable.tsv", show_col_types = FALSE)
rownames(coldata) <- coldata$sample

Warning: Setting row names on a tibble is deprecated.

coldata

counts <- read.csv("data/miR.Counts.csv", header = TRUE, sep = ",")
counts <- column_to_rownames(counts, var = "miRNA")
head(counts)

colnames(counts) <- gsub("^X", "", colnames(counts))

counts_samples <- colnames(counts)
phenotable_samples <- coldata$sample
common_samples <- intersect(counts_samples, phenotable_samples)

counts <- counts[, c(counts$miRNA, common_samples)]  
counts <- counts[, rownames(coldata)] #ранжирую по колонки в counts так же как и названия строк в coldata
head(counts)

anno <- read.csv("data/annotation.report.csv", header = TRUE, sep = ",")
anno$Sample.name.s. <- gsub("-", ".", anno$Sample.name.s.)
anno <- anno[, -c(2:5, 7, 15)]


common_samples <- intersect(anno$Sample.name.s., coldata$sample)

anno <- anno[anno$Sample.name.s. %in% common_samples, ]
anno <- anno[match(rownames(coldata), anno$Sample.name.s.), ] #ранжирую по колонки в counts так же как и названия строк в coldata
anno

Весь датасет

anno_long <- anno %>%
  pivot_longer(cols = -Sample.name.s., names_to = "RNA_Type", values_to = "Count")

plt <- ggplot(anno_long, aes(x = Sample.name.s., y = Count, fill = RNA_Type)) +
  geom_bar(stat = "identity") +
  theme_minimal() +
  labs(x = "Sample", y = "Read Count") +
  theme(axis.text.x = element_text(angle = 45, hjust = 1)) +
  scale_fill_brewer(palette = "Set3")  # Красивые цвета

print(plt)
ggsave("./pictures_transpl/transpl_barplot_alldataset_no_normalised.tiff", plot = plt, width = 8, height = 6, dpi = 300,  bg = "white")

anno_long <- anno %>%
  rowwise() %>%
  mutate(across(-Sample.name.s., ~ . / sum(c_across(-Sample.name.s.)))) %>% 
  ungroup() %>%
  pivot_longer(cols = -Sample.name.s., names_to = "RNA_Type", values_to = "Proportion")

plt <- ggplot(anno_long, aes(x = Sample.name.s., y = Proportion, fill = RNA_Type)) +
  geom_bar(stat = "identity") +
  theme_minimal() +
  labs(x = "Sample", y = "Proportion") +
  theme(axis.text.x = element_text(angle = 45, hjust = 1)) +
  scale_fill_brewer(palette = "Set3")

plt
ggsave("./pictures_transpl/transpl_barplot_alldataset_normalised.tiff", plot = plt, width = 8, height = 6, dpi = 300, bg = "white")

coldata$condition <- relevel(factor(coldata$condition), ref = "no_complications")
modelMatrix <- model.matrix(~condition, coldata)
modelMatrix

   (Intercept) conditioncellular conditionhumoral conditionTCAD
1            1                 0                0             0
2            1                 0                0             0
3            1                 0                1             0
4            1                 1                0             0
5            1                 0                0             1
6            1                 0                0             1
7            1                 0                0             1
8            1                 0                0             0
9            1                 0                0             0
10           1                 0                0             0
11           1                 0                0             0
12           1                 0                1             0
13           1                 0                1             0
14           1                 1                0             0
15           1                 1                0             0
16           1                 1                0             0
17           1                 1                0             0
18           1                 1                0             0
19           1                 0                0             1
20           1                 0                0             1
21           1                 0                1             0
22           1                 0                0             1
23           1                 0                0             1
24           1                 0                1             0
25           1                 0                1             0
26           1                 0                0             0
27           1                 0                0             0
28           1                 0                1             0
29           1                 0                0             0
attr(,"assign")
[1] 0 1 1 1
attr(,"contrasts")
attr(,"contrasts")$condition
[1] "contr.treatment"

qr(modelMatrix)$rank  # ранг матрицы

[1] 4

ncol(modelMatrix) 

[1] 4

dds <- DESeqDataSetFromMatrix(countData = counts, colData = coldata, design = ~ condition)

converting counts to integer mode

dds$condition <- relevel(dds$condition, ref = "no_complications")
dds

class: DESeqDataSet 
dim: 913 29 
metadata(1): version
assays(1): counts
rownames(913): Hsa-Let-7-P1a_3p* Hsa-Let-7-P1a_5p/P2a1_5p/P2a2_5p ... Hsa-Mir-9851_3p Hsa-Mir-9851_5p*
rowData names(0):
colnames(29): 105_S1_R1_001 197_S2_R1_001 ... 127_S29_R1_001 138_S30_R1_001
colData names(2): sample condition

dim(dds)

[1] 913  29

smallestGroupSize <- 15
keep <- rowSums(counts(dds) >= 10) >= smallestGroupSize
dds <- dds[keep,]
dim(dds)

[1] 297  29

Run Differential Expression Analysis

dds <- DESeq(dds, fitType = "parametric")

estimating size factors
estimating dispersions
gene-wise dispersion estimates
mean-dispersion relationship
final dispersion estimates
fitting model and testing
-- replacing outliers and refitting for 31 genes
-- DESeq argument 'minReplicatesForReplace' = 7 
-- original counts are preserved in counts(dds)
estimating dispersions
fitting model and testing

dds

class: DESeqDataSet 
dim: 297 29 
metadata(1): version
assays(6): counts mu ... replaceCounts replaceCooks
rownames(297): Hsa-Let-7-P1a_5p/P2a1_5p/P2a2_5p Hsa-Let-7-P1b_5p ... Hsa-Mir-96-P3_3p* Hsa-Mir-96-P3_5p
rowData names(31): baseMean baseVar ... maxCooks replace
colnames(29): 105_S1_R1_001 197_S2_R1_001 ... 127_S29_R1_001 138_S30_R1_001
colData names(4): sample condition sizeFactor replaceable

plotDispEsts(dds)

raw_counts <- counts(dds, normalized = FALSE)
normalized_counts <- counts(dds, normalized = TRUE)

df <- data.frame(
  Sample = rep(colnames(dds), 2),
  Counts = c(colSums(raw_counts), colSums(normalized_counts)),
  Type = rep(c("Raw", "Normalized"), each = ncol(dds))
)

plt <- ggplot(df, aes(x = Sample, y = Counts, fill = Type)) +
  geom_bar(stat = "identity", position = "dodge") +
  theme_minimal() +
  labs(title = "Counts before and after normalization", x = "Sample", y = "Total Counts") +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

plt
ggsave("./pictures_transpl/transpl_Counts before and after normalization.tiff", plot = plt, width = 8, height = 6, dpi = 300, bg = "white")

rlog трансформация

rlt <- rlog(dds)  #rlog Transformation
meanSdPlot(assay(rlt)) 

vsd <- varianceStabilizingTransformation(dds, blind=FALSE) 
meanSdPlot(assay(vsd)) #показывает, как изменяется стандартное отклонение в зависимости от среднего значения экспрессии

** PCA plot **

pcaData <- plotPCA(rlt, intgroup=c("condition"), returnData = TRUE)

using ntop=500 top features by variance

percentVar <- round(100 * attr(pcaData, "percentVar"))

pcaData$sample <- gsub("_.*", "", coldata$sample)


plt <- ggplot(pcaData, aes(PC1, PC2, color = condition)) +
  geom_text(aes(label=sample), size=3, vjust=1.5) +
  geom_point(size = 3) +
  xlab(paste0("PC1: ", percentVar[1], "%")) +
  ylab(paste0("PC2: ", percentVar[2], "%")) + 
  coord_fixed() +
  theme_bw() +
  scale_color_brewer(palette = "Set2")
plt
ggsave("./pictures_transpl/transpl_PCA plot.tiff", plot = plt, width = 8, height = 6, dpi = 300, bg = "white")

Plot a heatmap of 50 most expressed genes Этот heatmap отражает уровни экспрессии генов, а не разницу между группами. Цвета не означают up- или down-регуляцию в сравнении с контрольной группой, потому что heatmap показывает абсолютные значения экспрессии, а не fold change!

select <- order(rowMeans(counts(dds,normalized=TRUE)),
                decreasing=TRUE)[1:50]
df <- as.data.frame(colData(dds)$condition)
colnames(df) <- "condition"
rownames(df) <- colnames(counts(dds))
plt <- pheatmap(assay(rlt)[select,], 
         cluster_rows = TRUE, 
         show_rownames = TRUE, 
         cluster_cols = TRUE, 
         annotation_col = df,
         fontsize_row = 6) 

plt

ggsave("./pictures_transpl/transpl_Plot a heatmap of 50 most expressed genes.tiff", plot = plt, width = 8, height = 6, dpi = 300, bg = "white")

Plot of the distance between samples heatmap Расчет расстояний между образцами • Обычно используется евклидово расстояние (по умолчанию в DESeq2). • Оно вычисляется по нормализованным данным экспрессии (rlog() или vst()). • Чем меньше расстояние — тем более похожи образцы.

sampleDists <- dist(t(assay(rlt)))
sampleDistMatrix <- as.matrix(sampleDists)
rownames(sampleDistMatrix) <- paste(rlt$condition)
colnames(sampleDistMatrix) <- paste(rlt$condition)
colors <- colorRampPalette(rev(brewer.pal(9, "Blues")) )(255)

plt <- pheatmap(sampleDistMatrix,
         clustering_distance_rows = "euclidean",
         clustering_distance_cols = "euclidean",
         color = colors)

plt

ggsave("./pictures_transpl/transpl_Plot of the distance between samples heatmap.tiff", plot = plt, width = 8, height = 6, dpi = 300, bg = "white")

res_humoral <- results(dds, contrast=c("condition", "no_complications", "humoral"))
res_humoral

log2 fold change (MLE): condition no_complications vs humoral 
Wald test p-value: condition no_complications vs humoral 
DataFrame with 297 rows and 6 columns
                                    baseMean log2FoldChange     lfcSE      stat    pvalue      padj
                                   <numeric>      <numeric> <numeric> <numeric> <numeric> <numeric>
Hsa-Let-7-P1a_5p/P2a1_5p/P2a2_5p   72441.408      0.0618842  0.288818  0.214267 0.8303389  0.939409
Hsa-Let-7-P1b_5p                    2471.713     -0.9029022  0.495172 -1.823412 0.0682410  0.396094
Hsa-Let-7-P1c_5p                    1586.095     -1.0840729  0.471749 -2.297986 0.0215626  0.246039
Hsa-Let-7-P2a1_3p*                   106.535      0.0898204  0.728328  0.123324 0.9018504  0.974875
Hsa-Let-7-P2a2_3p*                   117.677     -2.1053617  1.182746 -1.780062 0.0750658  0.409032
...                                      ...            ...       ...       ...       ...       ...
Hsa-Mir-95-P3_5p                     11.6942       1.248042  1.169651  1.067020 0.2859626        NA
Hsa-Mir-96-P1_5p                    452.3189       1.016073  0.712354  1.426361 0.1537643 0.6037509
Hsa-Mir-96-P2_5p                 104065.5500      -0.309282  0.550995 -0.561315 0.5745829 0.8960804
Hsa-Mir-96-P3_3p*                    27.6400       0.953532  0.804588  1.185117 0.2359712 0.7233923
Hsa-Mir-96-P3_5p                   1433.3806       1.044344  0.364247  2.867132 0.0041421 0.0737293

MA plot Фильтрация точек с низким средним экспрессированием (по baseMean). • Обычно отсекаются baseMean < 1. 2. Определение значимых генов (синие точки): • Используется критерий padj < 0.1 по умолчанию, а не < 0.05!

tiff("./pictures_transpl/transpl_PlotMA_standart_padj_0.05_humoral.tiff", 
     width = 8, height = 6, units = "in", res = 300, bg = "white")
plotMA(res_humoral, alpha = 0.05, ylim = c(-8, 8)) 
dev.off()

null device 
          1 

plotMA(res_humoral, alpha = 0.05, ylim = c(-8, 8)) 

Кастомный MA plot

res_df <- res_humoral %>%
  as.data.frame() %>%
  mutate(color = case_when( 
    padj < 0.05  ~ "padj < 0.05",   
    pvalue < 0.05  ~ "pvalue < 0.05", 
    TRUE ~ "All"
  ))

plt <- ggplot(res_df, aes(x = baseMean, y = log2FoldChange, color = color)) +
  geom_point(alpha = 0.7, size = 1) +
  geom_hline(yintercept = 0, linetype = "solid", color = "gray40", size = 1.5) +
  scale_color_manual(values = c("All" = "gray70", 
                                "pvalue < 0.05" = "blue", 
                                "padj < 0.05" = "red")) +
  scale_x_log10(labels = scales::scientific) + 
  theme_minimal() +
  labs(x = "mean of normalized counts", 
       y = "log fold change", 
       color = NULL)

plt
ggsave("./pictures_transpl/transpl_Сustom MAplot_humoral.tiff", plot = plt, width = 8, height = 6, dpi = 300, bg = "white")

Значимые результаты

58 генов (20%) имеют низкие уровни экспрессии и фильтруются из анализа. independent filtering — процедура, которая исключает гены с низкими значениями для увеличения статистической мощности.

signres_humoral <- results(dds, contrast=c("condition", "no_complications", "humoral"), alpha=0.05) 
summary(signres_humoral)

out of 297 with nonzero total read count
adjusted p-value < 0.05
LFC > 0 (up)       : 0, 0%
LFC < 0 (down)     : 9, 3%
outliers [1]       : 1, 0.34%
low counts [2]     : 58, 20%
(mean count < 40)
[1] see 'cooksCutoff' argument of ?results
[2] see 'independentFiltering' argument of ?results

Let’s arranged it by log2FoldChange:

order_indices <- order(-res_humoral$log2FoldChange)
res_humoral[order_indices, ]

log2 fold change (MLE): condition no_complications vs humoral 
Wald test p-value: condition no_complications vs humoral 
DataFrame with 297 rows and 6 columns
                     baseMean log2FoldChange     lfcSE      stat      pvalue       padj
                    <numeric>      <numeric> <numeric> <numeric>   <numeric>  <numeric>
Hsa-Mir-542_3p       179.1170        1.55776  0.661653   2.35434   0.0185557   0.225199
Hsa-Mir-190-P1_5p     19.6983        1.54017  1.120054   1.37509   0.1691041         NA
Hsa-Mir-874_3p       183.9741        1.53509  0.798423   1.92265   0.0545243   0.378167
Hsa-Mir-624_5p        18.9670        1.36865  1.076852   1.27097   0.2037388         NA
Hsa-Mir-95-P3_5p      11.6942        1.24804  1.169651   1.06702   0.2859626         NA
...                       ...            ...       ...       ...         ...        ...
Hsa-Mir-873_3p       102.3872       -2.74537  1.229274  -2.23333 2.55274e-02 0.27263230
Hsa-Mir-10-P2a_5p   1162.4941       -2.85595  0.650038  -4.39350 1.11539e-05 0.00148904
Hsa-Mir-483_5p       144.9177       -3.86731  1.288810  -3.00068 2.69377e-03 0.06522233
Hsa-Mir-193-P1b_5p*  152.0458       -4.84885  0.982406  -4.93569 7.98689e-07 0.00021325
Hsa-Mir-193-P1b_3p    99.8971       -5.64893  1.644846  -3.43432 5.94044e-04 0.02643497

Visualisation for the first gene

#plotCounts(dds, gene=which.max(res_humoral$log2FoldChange), intgroup="condition")
plotCounts(dds, gene=which.min(res_humoral$padj), intgroup="condition")

#plotCounts(dds, gene=rownames(res)[which.min(res$padj[which.max(res$log2FoldChange)])], intgroup="condition")

Volcano plot

plt <- EnhancedVolcano(res_humoral,
                lab = rownames(res_humoral),
                x = "log2FoldChange",
                y = "padj",
                pCutoff = 0.05,
                FCcutoff = 1,
                labSize = 3.0,
                boxedLabels = FALSE,
                col = c('black', '#CBD5E8', '#B3E2CD', '#FDCDAC'),
                colAlpha = 1,
                title = NULL,        
                subtitle = NULL) 

plt
ggsave("./pictures_transpl/transpl_VolcanoPlot_humoral.tiff", plot = plt, width = 8, height = 6, dpi = 300, bg = "white")

coldata_filtered <- coldata[coldata$condition %in% c("humoral", "no_complications"), ]
coldata_filtered

Plot a heatmap of diff expressed genes

res_sign_humoral <- subset(res_humoral, padj < 0.05 & !is.na(padj) & abs(log2FoldChange) > 1.0)
res_sign_humoral <- res_sign_humoral[order(res_sign_humoral$log2FoldChange, decreasing = TRUE), ]

sig_genes <- rownames(res_sign_humoral)  

de_mat <- assay(rlt)[sig_genes, ] 

de_mat_filtered <- de_mat[, coldata_filtered$sample]
#datamatrix <- t(scale(t(de_mat_filtered)))
datamatrix <- de_mat_filtered

annotation_col <- data.frame(condition = coldata_filtered$condition)
rownames(annotation_col) <- colnames(datamatrix)

annotation_colors <- list(
  condition = c("no_complications" = "#FFCC00", "humoral" = "#3399FF"))

plt <- pheatmap(datamatrix,
         cluster_rows = TRUE, 
         show_rownames = TRUE, 
         cluster_cols = TRUE, 
         annotation_col = annotation_col,
         annotation_colors = annotation_colors,
         display_numbers = FALSE,
         legend = TRUE,
         fontsize = 15)  

ggsave("./pictures_transpl/transpl_Heatmap of diff expressed genes_humoral.tiff", plot = plt, width = 8, height = 6, dpi = 300, bg = "white")

Анализ обогащения гуморальный ответ

up_humoral <- res_sign_humoral %>% 
  as.data.frame() %>% 
  filter(log2FoldChange > 0)
down_humoral <- res_sign_humoral %>% 
  as.data.frame() %>% 
  filter(log2FoldChange < 0)
rownames(up_humoral)

character(0)

rownames(down_humoral)

[1] "Hsa-Mir-146-P1_5p"   "Hsa-Mir-425_3p*"     "Hsa-Mir-148-P2_3p"   "Hsa-Mir-10-P3c_5p"   "Hsa-Mir-10-P2a_5p"  
[6] "Hsa-Mir-193-P1b_5p*" "Hsa-Mir-193-P1b_3p" 

Переводим в miRBase • miRBase: https://www.mirbase.org/ • MirGeneDB: https://mirgenedb.org/

url <- "https://mirgenedb.org/browse/hsa"
page <- read_html(url)

Парсим таблицу

mir_table <- page %>%
  html_element("table") %>%
  html_table(fill = TRUE) 

mir_table <- mir_table[-c(1:3), c(1,2) ] 
colnames(mir_table) <- c("MirGeneDB_ID", "MiRBase_ID")
mir_table$MirGeneDB_ID <- sub(" V", "", mir_table$MirGeneDB_ID)

head(mir_table)

down_humoral_clean <- sub("_.*", "", row.names(down_humoral))
down_humoral_converted <- mir_table$MiRBase_ID[match(down_humoral_clean, mir_table$MirGeneDB_ID)]
down_humoral_converted

[1] "hsa-mir-146b"   "hsa-mir-425"    NA               "hsa-mir-125b-2" NA               "hsa-mir-193a"  
[7] "hsa-mir-193a"  

Конвертация в MIMATID NA без соответствия удалила из анализа

NA Hsa-Mir-148-P2_3p есть три похожих соответствия: Hsa-Mir-148-P1 hsa-mir-148a
Hsa-Mir-148-P3 hsa-mir-152
Hsa-Mir-148-P4 hsa-mir-148b

NA Hsa-Mir-10-P2a_5p есть три похожих соответствие: Hsa-Mir-10-P2b hsa-mir-99b
Hsa-Mir-10-P2c hsa-mir-99a
Hsa-Mir-10-P2d hsa-mir-100

[1] “Hsa-Mir-146-P1_5p” “Hsa-Mir-425_3p” ”Hsa-Mir-148-P2_3p” ”Hsa-Mir-10-P3c_5p” ”Hsa-Mir-10-P2a_5p”
[6] ”Hsa-Mir-193-P1b_5p” “Hsa-Mir-193-P1b_3p”

mirna_names_down <- c("hsa-miR-146b-5p", "hsa-miR-425-3p", "hsa-miR-125b-5p", "hsa-miR-193a-5p", "hsa-miR-193a-3p")
converted_mirna_down <- miRNAVersionConvert(mirna_names_down)
converted_mirna_down

Запрос таргетов из базы multiMiR

targets_humoral_down <- unique(get_multimir(org = "hsa", mirna = converted_mirna_down$Accession, table = "validated")@data$target_symbol)

Searching mirecords ...
Searching mirtarbase ...
Searching tarbase ...

#writeLines(targets_down, "targets_down150_list.txt")

Анализ обогащения из базы биологических процессов

#msig_go_bp <- msigdbr(species = "Homo sapiens", category = "C5", subcategory = "GO:BP")
# targets_down <- readLines("targets_down150_list.txt")
# targets_up <- readLines("targets_up150_list.txt")

GO_enrich_down_humoral_bp <- enrichGO(
  gene          = targets_humoral_down,  
  OrgDb         = org.Hs.eg.db,
  keyType       = "SYMBOL",
  ont           = "BP", 
  pAdjustMethod = "BH",
  qvalueCutoff  = 0.05
)

Визуализация

p1 <- dotplot(GO_enrich_down_humoral_bp, showCategory = 20) + 
  ggtitle("GO Enrichment for DOWNregulated targets") + 
  theme(
    plot.title = element_text(size = 12, face = "bold"),
    axis.text.y = element_text(size = 11)
  )

p1

ggsave("./pictures_transpl/transpl_GO_enrichment_dotplot_down_humoral_bp.tiff", plot = p1, width = 16, height = 10, dpi = 300)

GO_enrich_DOWN_humoral_BP <- enrichplot::pairwise_termsim(GO_enrich_down_humoral_bp, method = "JC")

plt <- emapplot(GO_enrich_DOWN_humoral_BP, 
         repel = TRUE,
         showCategory = 20) +
  ggtitle("Biological processes for DOWNregulated targets for humoral") +
  theme(
    plot.title = element_text(size = 12, face = "bold"),
    axis.text = element_text(size = 3)
    )    

plt

ggsave("./pictures_transpl/transpl_GO_enrichment_emapplot_DOWN_humoral_BP.tiff", plot = plt, width = 16, height = 10, dpi = 300)

GO_enrich_UP150_BP <- enrichplot::pairwise_termsim(GO_enrich_up150_bp, method = "JC")

plt <- emapplot(GO_enrich_UP150_BP, 
         repel = TRUE,
         showCategory = 20) +
  ggtitle("Biological processes for UPregulated targets for vesicles 150") +
  theme(
    plot.title = element_text(size = 12, face = "bold"),
    axis.text = element_text(size = 3)
    )   

plt

ggsave("./pictures/GO_enrichment_emapplot_BPup_type150.tiff", plot = plt, width = 16, height = 10, dpi = 300)

Анализ обогащения из базы IMMUNESIGDB IMMUNESIGDB: Наборы генов, связанные с иммунной системой, включая иммуно-онкологию и другие аспекты иммунитета.

msig_go_bp <- msigdbr(species = "Homo sapiens", category = "C7", subcategory = "IMMUNESIGDB")

GO_enrich_up150_immum <- enricher(gene = targets150_up, TERM2GENE = msig_go_bp[, c("gs_name", "gene_symbol")])
GO_enrich_down150_immun <- enricher(gene = targets150_down, TERM2GENE = msig_go_bp[, c("gs_name", "gene_symbol")])

Визуализация IMMUNESIGDB

p1 <- dotplot(GO_enrich_up150_immum, showCategory = 20) + 
  ggtitle("GO Enrichment IMMUNESIGD for UPregulated targets for vesicles 150") + 
  theme(
    plot.title = element_text(size = 12, face = "bold"),
    axis.text.y = element_text(size = 8)
  )

p2 <- dotplot(GO_enrich_down150_immun, showCategory = 20) + 
  ggtitle("GO Enrichment IMMUNESIGD for DOWNregulated targets for vesicles 150") + 
  theme(
    plot.title = element_text(size = 12, face = "bold"),
    axis.text.y = element_text(size = 8)
  )

p1 + p2

combined_plot <- p1 + p2

ggsave("./pictures/GO_enrichment_dotplot_IMMUNESIGD_type150.tiff", plot = combined_plot, width = 16, height = 10, dpi = 300)

Анализ обогащения из базы KEGG

msig_go_bp <- msigdbr(species = "Homo sapiens", category = "C2", subcategory = "CP:KEGG")

GO_enrich_up150_KEGG <- enricher(gene = targets150_up, TERM2GENE = msig_go_bp[, c("gs_name", "gene_symbol")])
GO_enrich_down150_KEGG <- enricher(gene = targets150_down, TERM2GENE = msig_go_bp[, c("gs_name", "gene_symbol")])

p1 <- dotplot(GO_enrich_up150_KEGG, showCategory = 20) + 
  ggtitle("GO Enrichment KEGG for UPregulated targets for vesicles 150") + 
  theme(
    plot.title = element_text(size = 12, face = "bold"),
    axis.text.y = element_text(size = 8)
  )

p2 <- dotplot(GO_enrich_down150_KEGG, showCategory = 20) + 
  ggtitle("GO Enrichment KEGG for DOWNregulated targets for vesicles 150") + 
  theme(
    plot.title = element_text(size = 12, face = "bold"),
    axis.text.y = element_text(size = 8)
  )

p1 + p2

combined_plot <- p1 + p2

ggsave("./pictures/GO_enrichment_dotplot_KEGG_type150.tiff", plot = combined_plot, width = 16, height = 10, dpi = 300)

Анализ обогащения из базы CP:WIKIPATHWAYS

msig_go_bp <- msigdbr(species = "Homo sapiens", category = "C2", subcategory = "CP:WIKIPATHWAYS")

GO_enrich_up150_WIKI <- enricher(gene = targets150_up, TERM2GENE = msig_go_bp[, c("gs_name", "gene_symbol")])
GO_enrich_down150_WIKI <- enricher(gene = targets150_down, TERM2GENE = msig_go_bp[, c("gs_name", "gene_symbol")])

p1 <- dotplot(GO_enrich_up150_WIKI, showCategory = 20) + 
  ggtitle("GO Enrichment WIKIPATHWAYS for UPregulated targets for vesicles 150") + 
  theme(
    plot.title = element_text(size = 12, face = "bold"),
    axis.text.y = element_text(size = 8)
  )

p2 <- dotplot(GO_enrich_down150_WIKI, showCategory = 20) + 
  ggtitle("GO Enrichment WIKIPATHWAYS for DOWNregulated targets for vesicles 150") + 
  theme(
    plot.title = element_text(size = 12, face = "bold"),
    axis.text.y = element_text(size = 8)
  )

p1 + p2
combined_plot <- p1 + p2

ggsave("./pictures/GO_enrichment_dotplot_WIKIPATHWAY_type150.tiff", plot = combined_plot, width = 16, height = 10, dpi = 300)

Только везикулы типа 16

coldata_16 <- coldata[coldata$type == 16, ]
rownames(coldata_16) <- coldata_16$sample
coldata_16

common_samples_16 <- intersect(colnames(counts), coldata_16$samples)

counts_16 <- counts[, c(counts$miRNA, common_samples)]  
counts_16 <- counts_16[, rownames(coldata_16)] #ранжирую по колонки в counts так же как и названия строк в coldata_150
head(counts_16)

Создаем DESeqDataSet из матрицы каунтов

dds_16 <- DESeqDataSetFromMatrix(countData = counts_16, 
                              colData = coldata_16, 
                              design = ~ 0 + patient + condition)

converting counts to integer mode

dds_16$condition <- relevel(dds_16$condition, ref = "before")
dds_16

class: DESeqDataSet 
dim: 913 6 
metadata(1): version
assays(1): counts
rownames(913): Hsa-Let-7-P1a_3p* Hsa-Let-7-P1a_5p/P2a1_5p/P2a2_5p ... Hsa-Mir-9851_3p Hsa-Mir-9851_5p*
rowData names(0):
colnames(6): 29.1p16_S39_R1_001 15.1p16_S33_R1_001 ... 15.7p16_S31_R1_001 17.7p16_S37_R1_001
colData names(4): sample condition type patient

Фильтрация

dim(dds_16)

[1] 913   6

smallestGroupSize <- 3
keep <- rowSums(counts(dds_16) >= 10) >= smallestGroupSize
dds_16 <- dds_16[keep,]
dim(dds_16)

[1] 219   6

Run Differential Expression Analysis for 150 type

dds_16 <- DESeq(dds_16, fitType = "parametric")

estimating size factors
estimating dispersions
gene-wise dispersion estimates
mean-dispersion relationship
final dispersion estimates
fitting model and testing

dds_16

class: DESeqDataSet 
dim: 219 6 
metadata(1): version
assays(4): counts mu H cooks
rownames(219): Hsa-Let-7-P1a_5p/P2a1_5p/P2a2_5p Hsa-Let-7-P1b_5p ... Hsa-Mir-96-P2_5p Hsa-Mir-96-P3_5p
rowData names(30): baseMean baseVar ... deviance maxCooks
colnames(6): 29.1p16_S39_R1_001 15.1p16_S33_R1_001 ... 15.7p16_S31_R1_001 17.7p16_S37_R1_001
colData names(5): sample condition type patient sizeFactor

plotDispEsts(dds_16)

res_16 <- results(dds_16, contrast=c("condition", "before", "after"))
res_16

log2 fold change (MLE): condition before vs after 
Wald test p-value: condition before vs after 
DataFrame with 219 rows and 6 columns
                                   baseMean log2FoldChange     lfcSE      stat    pvalue      padj
                                  <numeric>      <numeric> <numeric> <numeric> <numeric> <numeric>
Hsa-Let-7-P1a_5p/P2a1_5p/P2a2_5p 22993.7673      0.1901164  0.224536  0.846707 0.3971586  0.926559
Hsa-Let-7-P1b_5p                  2527.3696      0.4961225  0.280481  1.768830 0.0769222  0.575168
Hsa-Let-7-P1c_5p                   481.2148      0.5330152  0.461003  1.156209 0.2475958  0.860794
Hsa-Let-7-P2a1_3p*                  23.7282      0.4135740  1.966411  0.210319 0.8334185  0.954993
Hsa-Let-7-P2a3_5p                45006.2426     -0.0642043  0.238029 -0.269733 0.7873659  0.940503
...                                     ...            ...       ...       ...       ...       ...
Hsa-Mir-92-P2c_5p*                  25.3899      -0.754468  2.275831 -0.331513 0.7402570  0.931703
Hsa-Mir-95-P2_3p                   193.6126      -0.340069  0.662909 -0.512995 0.6079550  0.926559
Hsa-Mir-96-P1_5p                    14.4204      -4.216533  2.813830 -1.498503 0.1340025  0.764653
Hsa-Mir-96-P2_5p                  4249.9579      -0.537344  0.268063 -2.004539 0.0450123  0.487244
Hsa-Mir-96-P3_5p                   121.7907       0.504262  0.806362  0.625355 0.5317382  0.926559

MA plot

Фильтрация точек с низким средним экспрессированием (по baseMean). • Обычно отсекаются baseMean < 1. 2. Определение значимых генов (синие точки): • Используется критерий padj < 0.1 по умолчанию, а не < 0.05!

tiff("./pictures/PlotMA_standart_padj_0.05_type16.tiff", 
     width = 8, height = 6, units = "in", res = 300, bg = "white")
plotMA(res_16, alpha = 0.05, ylim = c(-8, 8)) 
dev.off()

null device 
          1 

plotMA(res_16, alpha = 0.05, ylim = c(-8, 8)) 

Кастомный MA plot по p-value

res_df <- res_16 %>%
  as.data.frame %>%
  mutate(color = case_when( 
    pvalue < 0.05 & !is.na(pvalue) & abs(log2FoldChange) > 1  ~ "blue",  # Значимые по p-value и диф экспрессированные
    TRUE ~ "gray70"
  ))

plt <- ggplot(res_df, aes(x = baseMean, y = log2FoldChange, color = color)) +
  geom_point(alpha = 0.7, size = 1) +
  geom_hline(yintercept = 0, linetype = "solid", color = "gray40", size = 1.5) +  # Добавляем линию
  scale_color_manual(values = c("gray70" = "gray70", "blue" = "blue")) +
  scale_x_log10(labels = scales::scientific) + 
  theme_minimal() +
  labs(x = "mean of normalized counts", y = "log fold change") +
  theme(legend.position = "none")

plt
ggsave("./pictures/PlotMA_castom_pvalue0.05_type16.tiff", plot = plt, width = 8, height = 6, dpi = 300, bg = "white")

Значимые результаты

summary(results(dds_16, contrast=c("condition", "before", "after"), alpha=0.05))

out of 219 with nonzero total read count
adjusted p-value < 0.05
LFC > 0 (up)       : 0, 0%
LFC < 0 (down)     : 0, 0%
outliers [1]       : 0, 0%
low counts [2]     : 0, 0%
(mean count < 8)
[1] see 'cooksCutoff' argument of ?results
[2] see 'independentFiltering' argument of ?results

Let’s arranged it by log2FoldChange:

order_indices <- order(-res_16$log2FoldChange)
res_16[order_indices, ]

log2 fold change (MLE): condition before vs after 
Wald test p-value: condition before vs after 
DataFrame with 219 rows and 6 columns
                    baseMean log2FoldChange     lfcSE      stat     pvalue      padj
                   <numeric>      <numeric> <numeric> <numeric>  <numeric> <numeric>
Hsa-Mir-185_5p       14.8310        5.85950   2.69426   2.17481 0.02964445  0.487244
Hsa-Mir-136_5p*      13.3826        5.83711   3.16104   1.84658 0.06480773  0.575168
Hsa-Mir-197_3p       45.5597        4.16951   1.59282   2.61769 0.00885267  0.357073
Hsa-Mir-101-P1_5p*   16.5354        4.08972   2.30004   1.77811 0.07538608  0.575168
Hsa-Mir-10-P3a_5p    37.9479        3.82488   1.69051   2.26256 0.02366292  0.487244
...                      ...            ...       ...       ...        ...       ...
Hsa-Mir-154-P17_3p   15.3729       -4.07531   2.60884  -1.56211  0.1182617  0.764653
Hsa-Mir-190-P1_5p    10.5910       -4.10593   3.44201  -1.19289  0.2329137  0.850135
Hsa-Mir-96-P1_5p     14.4204       -4.21653   2.81383  -1.49850  0.1340025  0.764653
Hsa-Mir-128-P1_5p*   15.3208       -4.49032   3.05489  -1.46988  0.1415942  0.764653
Hsa-Mir-197_5p*      17.0795       -5.15605   2.93332  -1.75775  0.0787901  0.575168

Visualisation for the first gene

plotCounts(dds_16, gene=which.max(res_16$log2FoldChange), intgroup="condition")

plotCounts(dds_16, gene=which.min(res_16$pvalue), intgroup="condition")

#plotCounts(dds, gene=rownames(res)[which.min(res$padj[which.max(res$log2FoldChange)])], intgroup="condition")

Volcano plot

plt <- EnhancedVolcano(res_16,
                lab = rownames(res_16),
                x = "log2FoldChange",
                y = "pvalue",
                pCutoff = 0.05,
                FCcutoff = 1,
                labSize = 3.0,
                boxedLabels = FALSE,
                col = c('black', '#CBD5E8', '#B3E2CD', '#FDCDAC'),
                colAlpha = 1,
                title = NULL,        
                subtitle = NULL) 

plt
ggsave("./pictures/Volcano plot_based_on_Pvalue_16type.tiff", plot = plt, width = 8, height = 6, dpi = 300, bg = "white")

rlog трансформация • Черные точки – стандартное отклонение отдельных генов. • Красная линия – сглаженный тренд зависимости SD от среднего значения экспрессии. • Если красная линия наклонена вверх → стандартное отклонение растёт с увеличением среднего (плохая нормализация). • Если красная линия примерно горизонтальна → нормализация сработала хорошо.

rlt_16 <- rlog(dds_16) 
meanSdPlot(assay(rlt_16))  #показывает, как изменяется стандартное отклонение в зависимости от среднего значения экспрессии.

pcaData <- plotPCA(rlt_16, intgroup=c("condition", "patient"), returnData = TRUE)

using ntop=500 top features by variance

percentVar <- round(100 * attr(pcaData, "percentVar"))

ggplot(pcaData, aes(PC1, PC2, shape = patient, color = condition)) +
  geom_point(size = 3) +
  xlab(paste0("PC1: ", percentVar[1], "%")) +
  ylab(paste0("PC2: ", percentVar[2], "%")) + 
  coord_fixed() +
  theme_bw() +
  scale_color_brewer(palette = "Set2")

assay(rlt_16) <- limma::removeBatchEffect(assay(rlt_16),
                                       batch = colData(dds_16)[,'patient'])

pcaData <- plotPCA(rlt_16, intgroup=c("condition", "patient"), returnData = TRUE)

using ntop=500 top features by variance

percentVar <- round(100 * attr(pcaData, "percentVar"))

plt <- ggplot(pcaData, aes(PC1, PC2, shape = patient, color = condition)) +
  geom_point(size = 3) +
  xlab(paste0("PC1: ", percentVar[1], "%")) +
  ylab(paste0("PC2: ", percentVar[2], "%")) + 
  coord_fixed() +
  theme_bw() +
  scale_color_brewer(palette = "Set2")

plt
ggsave("./pictures/PCA plot for type 16 after removing donor effect.tiff", plot = plt, width = 8, height = 6, dpi = 300, bg = "white")

Plot a heatmap of the most expressed genes

res_sign_16 <- subset(res_16, pvalue < 0.05 & !is.na(pvalue) & abs(log2FoldChange) > 1.0)
res_sign_16 <- res_sign_16[order(res_sign_16$log2FoldChange, decreasing = TRUE), ]

sig_genes <- rownames(res_sign_16)  # Получаем имена генов, которые прошли фильтрацию

de_mat <- assay(rlt_16)[sig_genes, ] 
datamatrix <- t(scale(t(de_mat)))

annotation_col <- data.frame(condition = coldata_16$condition)
rownames(annotation_col) <- colnames(datamatrix)

annotation_colors <- list(
  condition = c("before" = "#FFCC00", "after" = "#3399FF")
)

plt <- pheatmap(datamatrix, 
         cluster_rows = TRUE, 
         show_rownames = TRUE, 
         cluster_cols = TRUE, 
         annotation_col = annotation_col,
         annotation_colors = annotation_colors,
         display_numbers = TRUE,
         legend = FALSE,
         fontsize = 15)  

plt

ggsave("./pictures/Heatmap of diff expressed genes_type16.tiff", plot = plt, width = 8, height = 6, dpi = 300, bg = "white")

Plot of the distance between samples heatmap Расчет расстояний между образцами • Обычно используется евклидово расстояние (по умолчанию в DESeq2). • Оно вычисляется по нормализованным данным экспрессии (rlog() или vst()). • Чем меньше расстояние — тем более похожи образцы.

sampleDists_16 <- dist(t(assay(rlt_16)))
sampleDistMatrix_16 <- as.matrix(sampleDists_16)
rownames(sampleDistMatrix_16) <- paste(rlt_16$condition, rlt_16$patient, sep="_patient")
colnames(sampleDistMatrix_16) <- paste(rlt_16$condition, rlt_16$patient, sep="_patient")

plt <- pheatmap(sampleDistMatrix_16,
         clustering_distance_rows = "euclidean",
         clustering_distance_cols = "euclidean",
         fontsize = 12,
         legend = FALSE,
         display_numbers = TRUE,
         color = colors)

plt

ggsave("./pictures/Plot of the distance between samples_type16.tiff", plot = plt, width = 8, height = 6, dpi = 300, bg = "white")

Анализ обогащения для везикул типа 16

up_16 <- res_sign_16 %>% 
  as.data.frame() %>% 
  filter(log2FoldChange > 0)
down_16 <- res_sign_16 %>% 
  as.data.frame() %>% 
  filter(log2FoldChange < 0)
rownames(up_16)

[1] "Hsa-Mir-185_5p"     "Hsa-Mir-197_3p"     "Hsa-Mir-10-P3a_5p"  "Hsa-Mir-148-P3_5p*" "Hsa-Mir-30-P1a_3p*" "Hsa-Mir-150_3p*"   
[7] "Hsa-Mir-361_3p*"    "Hsa-Mir-10-P2b_5p"  "Hsa-Mir-126_3p*"   

rownames(down_16)

[1] "Hsa-Mir-340_5p"     "Hsa-Mir-181-P2c_5p" "Hsa-Mir-223_5p*"    "Hsa-Mir-30-P1c_3p*"

Переводим в miRBase • miRBase: https://www.mirbase.org/ • MirGeneDB: https://mirgenedb.org/

url <- "https://mirgenedb.org/browse/hsa"
page <- read_html(url)

Парсим таблицу

mir_table <- page %>%
  html_element("table") %>%
  html_table(fill = TRUE) 

mir_table <- mir_table[-c(1:3), c(1,2) ] 
colnames(mir_table) <- c("MirGeneDB_ID", "MiRBase_ID")
mir_table$MirGeneDB_ID <- sub(" V", "", mir_table$MirGeneDB_ID)

head(mir_table)

up_16_clean <- sub("_.*", "", row.names(up_16))
up_16_converted <- mir_table$MiRBase_ID[match(up_16_clean, mir_table$MirGeneDB_ID)]

#down_16_clean <- sub("_.*", "", row.names(down_16))
down_16_converted <- mir_table$MiRBase_ID[match(down_16_clean, mir_table$MirGeneDB_ID)]
up_16_converted

[1] "hsa-mir-185" "hsa-mir-197" NA            "hsa-mir-152" "hsa-mir-30d" "hsa-mir-150" "hsa-mir-361" "hsa-mir-99b" NA           

down_16_converted

[1] "hsa-mir-340"  "hsa-mir-181d" "hsa-mir-223"  NA            

Конвертация в MIMATID в итоге заменила NA вручную на самые близкие, но это такая себе практика

NA Hsa-Mir-10-P3a_5p есть два соответствия: Hsa-Mir-10-P1c = hsa-mir-10a Hsa-Mir-10-P3b = hsa-mir-125a

NA Hsa-Mir-126_3p* есть одно соответствие: Hsa-Mir-126-P2 = hsa-mir-126

NA Hsa-Mir-30-P1c_3p* есть три соответствия: Hsa-Mir-30-P1a = hsa-mir-30d Hsa-Mir-30-P1b = hsa-mir-30a Hsa-Mir-30-P1d = hsa-mir-30e

[1] “Hsa-Mir-185_5p” “Hsa-Mir-197_3p” “Hsa-Mir-10-P3a_5p” “Hsa-Mir-148-P3_5p” ”Hsa-Mir-30-P1a_3p” “Hsa-Mir-150_3p”
[7] ”Hsa-Mir-361_3p” “Hsa-Mir-10-P2b_5p” “Hsa-Mir-126_3p”
[1] ”Hsa-Mir-340_5p” ”Hsa-Mir-181-P2c_5p” ”Hsa-Mir-223_5p” “Hsa-Mir-30-P1c_3p*”

MI (MicroRNA Gene ID) — это идентификатор предшественника (precursor) miRNA MIMAT (Mature miRNA ID) — это идентификатор зрелой (mature) miRNA, которая функционирует в клетке

up_16_converted <-  c("hsa-mir-185-5p", "hsa-mir-197-3p", "hsa-mir-152-5p", "hsa-mir-30d-3p", "hsa-mir-150-3p", "hsa-mir-361-3p", "hsa-mir-99b-5p", "hsa-mir-126-3p")
down_16_converted <- c("hsa-mir-340-5p", "hsa-mir-181d-5p", "hsa-mir-223-5p")

# up_16_converted <-  c("hsa-mir-185", "hsa-mir-197", "hsa-mir-152", "hsa-mir-30d", "hsa-mir-150", "hsa-mir-361", "hsa-mir-99b", "hsa-mir-126")
# down_16_converted <- c("hsa-mir-340", "hsa-mir-181d", "hsa-mir-223") #но есть только MI, а не MIMAT

converted_mirna_up16 <- miRNAVersionConvert(up_16_converted)
converted_mirna_down16 <- miRNAVersionConvert(down_16_converted)
converted_mirna_up16

converted_mirna_down16

Запрос таргетов из базы multiMiR

targets16_up <- unique(get_multimir(org = "hsa", mirna = converted_mirna_up16$Accession, table = "predicted")@data$target_symbol)
targets16_up

targets16_down <- unique(get_multimir(org = "hsa", mirna = converted_mirna_down16$Accession, table = "predicted")@data$target_symbol)
targets16_down